[Ontology-editors] haspart documentation

[Chris Mungall]

On Jun 17, 2009, at 1:50 PM, Karen Christie wrote:

> Hi,
>
> Working my way through Midori's example and the logic for reasoning,  
> I now have a question on the reasoning for part_of.
>
> So, this one is fine:
>
> If child term A (mitochondrion) is a subclass of term B (intracellular
> organelle), and term B (intracellular organelle) is part of term C
> (cell), term A (mitochondrion) is part of term C (cell).

Yes

It's easier if you omit "term" from the above

>
> My question is on this one:
>
> Term A (mitochondrial membrane) part of term B (mitochondrion), and
> term B (mitochondrion) is a term A (intracellular organelle),
> therefore Term A (mitochondrial membrane) is part of term C
> (intracellular organelle)

yes

>
> It seems odd to me, to describe a thing that actually exists in cells,
> the mitochondrial membrane (term A) as being part of something that is
> a grouping term for a variety of cellular structures. However, this
> seems OK logically if it means that mitochondrial membrane is part of
> AN intracellular organelle.

This is exactly what the type level part_of relation means (if you  
stick an "all" before your statement above)

X (type level) part_of Y :
	ALL instances of X are part_of SOME instance of Y

> However, if I put a slightly different word 'ALL' in place of 'AN',
> then this isn't true:
> mitochondrial membrane is part of ALL intracellular organelle.
>
> So, I just wanted some clarification about how to think about the  
> logic.

I think it's important that the documentation sticks as closely as  
possible to the terminology of the OBO Relations ontology and it's  
accompanying paper:

	http://genomebiology.com/2005/6/5/R46

This makes it absolutely explicit (and it describes in formal terms  
what GO has always done, even preceding this paper)

In particular

- I would avoid using "the" as it confuses things. E.g. "the  
mitochondrial membrane" sounds like a particular MM.
- part_of, has_part, etc all follow the ALL-SOME pattern.



>
> thanks,
>
> -Karen
>
>
>
>
> On Wed, 17 Jun 2009, Midori Harris wrote:
>
>>> - good examples for the has part transitivity rules
>>
>> As promised yesterday, here's an example drawn from SF 2784316,  
>> which motivated the push to get has_part live.
>>
>> Briefly:
>> 'something' = precatalytic spliceosome
>> 'something else' = U4/U6 x U5 tri-snRNP complex
>> 'another thing' = snRNP U5
>>
>> More detail:
>> At present we have (greatly oversimplified):
>>
>> spliceosome GO:0005681
>> [i] U2-dependent spliceosome GO:0005684
>> --[p] snRNP U5 GO:0005682
>> --[p] [several other snRNP U(n) terms]
>>
>> Or, as a sentence, "snRNP U5 part_of U2-dependent spliceosome is_a  
>> spliceosome".
>>
>> We also have another CC term, 'U4/U6 x U5 tri-snRNP complex' (GO: 
>> 0046540), which is is_a snRNP and has no direct part_of parent.
>>
>> But it's now known that there are several more types of  
>> spliceosome, not all of which are catalytically active, and many of  
>> which are formed sequentially by the addition, removal, and  
>> rearrangement of various snRNPs. A given snRNP, such as snRNP U5,  
>> can thus be part of any of several different spliceosomal  
>> complexes, and the part_of relationship now in GO is wrong.
>>
>> In the has_part pilot project, the oversimplified example will  
>> become:
>>
>> precatalytic spliceosome GO:new
>> --[hp] U4/U6 x U5 tri-snRNP complex GO:0046540
>> ----[hp] snRNP U5 GO:0005682
>>
>> "Precatalytic spliceosome has_part U4/U6 x U5 tri-snRNP complex  
>> has_part snRNP U5".
>>
>> Thinking about this example, I now doubt that has_part is  
>> transitive over is_a, though. The new term will be is_a spliceosome:
>>
>> spliceosomal complex GO:0005681 [renamed]
>> [i] U2-dependent spliceosome GO:0005684
>> --[i] precatalytic spliceosome GO:new
>>
>> ... and not all spliceosomal complexes have the U4/U6 x U5 tri- 
>> snRNP complex, or even just U5. So I don't think the has_part.is_a  
>> section has the correct conclusion. (I think is_a is transitive  
>> over has_part, i.e. the is_a.has_part section is OK.)
>>
>> hope this is useful ...
>>
>> m
>>
>> _______________________________________________
>> Ontology-editors mailing list
>> Ontology-editors at geneontology.org
>> http://fafner.stanford.edu/mailman/listinfo/ontology-editors
>>
> _______________________________________________
> Ontology-editors mailing list
> Ontology-editors at geneontology.org
> http://fafner.stanford.edu/mailman/listinfo/ontology-editors
>